I/O Issues with mergerfs

[Jacob]

Currently running the bulk of my storage through a mergerfs/snapraid pool and have two other drives outside of that pool for various other things. The issue I am running into is that I want to create a virtio drive for a VM that I want located on the pool because it has more storage. I tried moving the existing drive onto the pool but I get an error:

update VM 100: -virtio0 CT_VM_Backups:600,format=qcow2,backup=0
TASK ERROR: error with cfs lock 'storage-CT_VM_Backups': unable to create image: got lock timeout - aborting command

When I tried to manually move the drive file from it's location outside of the pool to the storage location inside the pool, it would not register in the VM. Then when I created a smaller (32GB) virtual drive that was located inside the pool and wanted to format that drive inside the VM, I would only get I/O errors and the system wouldn't create or modify the drive file at all.

Would this indicate that one of my drives is dieing? I checked the individual drives and only one of them gives me some cause for concern but the directory I am trying to write too is not located on this particular drive. I've been able to complete backup operations and move large files in and out of the mergerfs pool without issue. Not sure why creating this virtual drive is giving me such issue.

 

[fiona]

Hi,
did you manage to resolve the issue in the meantime?

Currently running the bulk of my storage through a mergerfs/snapraid pool and have two other drives outside of that pool for various other things. The issue I am running into is that I want to create a virtio drive for a VM that I want located on the pool because it has more storage. I tried moving the existing drive onto the pool but I get an error:

update VM 100: -virtio0 CT_VM_Backups:600,format=qcow2,backup=0
TASK ERROR: error with cfs lock 'storage-CT_VM_Backups': unable to create image: got lock timeout - aborting command

This either means that the lock is held by something else (any other operations involving virtual disks running at the same time?) or that the storage itself is hanging.

When I tried to manually move the drive file from it's location outside of the pool to the storage location inside the pool, it would not register in the VM.

If nothing else works you could manually add a line for the drive to the VM config, but it's not recommended. First move the file to 100/vm-100-disk-<ID>.qcow2 where ID is so that it's different from the existing drives, add

unused0: CT_VM_Backups:100/vm-100-disk-<ID>.qcow2

to the config of the VM, i.e. /etc/pve/qemu-server/100.conf and then attach it to the VM in the GUI.

Then when I created a smaller (32GB) virtual drive that was located inside the pool and wanted to format that drive inside the VM, I would only get I/O errors and the system wouldn't create or modify the drive file at all.

So the timeout only happens when the virtual drive you want to add is large?

Would this indicate that one of my drives is dieing? I checked the individual drives and only one of them gives me some cause for concern but the directory I am trying to write too is not located on this particular drive. I've been able to complete backup operations and move large files in and out of the mergerfs pool without issue. Not sure why creating this virtual drive is giving me such issue.

 

[Jacob]

I did not resolve the issue. I ended up adding more SSD storage to my LVM and then moved over the data there. I don't know if the issue is due to the size, or if it's trying to duel allocate the resources to the drive pool. I'll try your suggestion and see if I am able to add a drive.